Feasibility of functional precision medicine for guiding treatment of relapsed or refractory pediatric cancers

Cancer remains the leading cause of disease-related death in children and teenagers in the United States. While survival has improved for some cancers, high-risk, relapsed, or refractory pediatric cancers have limited established treatment options. Genomics-guided precision oncology seeks to match treatments to tumor molecular alterations, and multiple programs (e.g., Zero Childhood Cancer in Australia, PROFYLE in Canada, iTHER in the Netherlands) have shown benefit. However, genomics alone is constrained in many pediatric cancers, which often lack actionable driver mutations or are driven by copy number alterations and fusions. To overcome these limitations, integrating ex vivo drug sensitivity testing (DST) with genomic profiling—functional precision medicine (FPM)—can expand therapeutic options by testing patient-derived tumor cells directly against approved drugs. The study’s aim was to determine the feasibility of generating and returning FPM data (DST plus targeted genomic profiling) for pediatric patients with relapsed or refractory cancers in a clinically actionable timeframe and to explore clinical outcomes when FPM guided treatment decisions.

Prior pediatric precision oncology initiatives using sequencing have identified molecular alterations but frequently lacked matched, actionable therapies, especially in pediatric tumors driven by copy number changes or fusions. The INFORM study in Europe began integrating functional DST to supplement genomics, providing additional options when genomics alone was insufficient. Adult interventional FPM trials in Finland and Austria for hematological malignancies demonstrated that combining molecular profiling with high-throughput DST can provide clinical benefit. However, prior FPM trials had enrolled exclusively adults and largely focused on hematological cancers, reflecting technical challenges of DST in solid tumors. Prospective FPM studies for pediatric patients were lacking, highlighting the need addressed by this study.

Study design: Prospective, non-randomized, single-arm observational feasibility study (ClinicalTrials.gov: NCT03860376) conducted at Nicklaus Children’s Hospital (Miami, Florida, USA), enrolling from 21 February 2019 to 31 December 2022. Inclusion: patients ≤21 years with suspected/confirmed recurrent or refractory cancer, planned/recent biopsy or excision (solid) or bone marrow aspirate (hematologic), and willingness for germline sample. Exclusion: no accessible malignant tissue, insufficient tissue for DST/genomics, newly diagnosed tumors or those with >90% cure rate on safe standard therapy. Primary endpoint: return of actionable FPM-based treatment recommendation(s) (DST and/or genomics) to the FPM tumor board (FPMTB) within 4 weeks, targeting ≥60% of enrolled patients. Secondary endpoints: objective response rate (ORR), progression-free survival (PFS) versus prior regimen and versus treatment of physician’s choice (TPC), and PFS ratio (PFS2/PFS1) ≥1.3. Patient cohort and samples: 25 pediatric/adolescent patients (19 solid, 6 hematologic) after exhausting standard-of-care. Solid tumor biopsies/resections (n=18 total; 1 biopsy, 17 resections) and hematologic cancer samples (n=6) were processed for ex vivo DST and genomic panel profiling (UCSF500) as tissue allowed. Median time from clinic to lab arrival was <48 h. Drug sensitivity testing (DST): Short-term patient-derived cultures (PDCs) were established via mechanical and enzymatic dissociation (Liberase DH + DNase I), RBC lysis, and culture in appropriate media; hematologic mononuclear cells were isolated by Ficoll gradient and cultured in mononuclear cell medium. PDCs were seeded at 1,000 cells/well in 384-well plates. A custom library of up to 125 FDA-approved and late-phase agents (including pediatric formulary/non-formulary anti-cancer drugs and selected repurposing candidates) was tested in duplicate across ten concentrations (10 µM–0.5 nM) for 72 h. Viability was measured via CellTiter-Glo (hematologic) or CellTiter-Glo 3D (solid). Dose-response data were used to calculate drug sensitivity scores (DSS; AUC-based) and IC50. Drugs were ranked by DSS and considered recommendable if IC50 ≤ clinically achievable plasma concentration (Cmax) from human PK data. Physician-requested drug combinations were tested when material allowed. DST plate quality control used Z-prime scores from positive/negative controls; plates passing criteria were analyzed. Repeatability of DSS and IC50 was assessed in selected repeats. Median PDC viability at testing was reported. Molecular profiling: Tumor panel sequencing using UCSF500 on FFPE tumor with matched blood (solids) or whole blood/buccal swab (hematologic) when tissue allowed. Some patients also had results from Foundation Medicine or CHLA OncoKids; available results were included. Additional validations included RT-qPCR confirming transcript losses and RNA-seq deconvolution confirming cell composition and tumor purity; PDCs’ genomic alterations were compared with original tumor profiles using WES/WTS data. FPM tumor board (FPMTB): Multidisciplinary team (treating physicians, pharmacists, nurses, precision medicine specialists, research coordinators, and translational researchers) reviewed DST and genomics results as they became available, considering drug availability, off-label access, insurance, prior treatments, and physician expertise to provide prioritized therapeutic options with recommended doses/schedules. Treatments were not administered per protocol; treating physicians made final decisions under separate consent. Statistics: Primary outcome analyzed with one-sided exact binomial test. Between-cohort PFS comparisons via two-sample logrank. Within-patient previous vs study regimen PFS via Cox regression with clustered computation. PFS ratios analyzed via Wilcoxon matched pairs; incidence of PFS ratio ≥1.3 via Barnard’s test. Additional tests included Wilcoxon, Kruskal–Wallis, McNemar, Chi-square, Spearman, and Kolmogorov–Smirnov as appropriate. Receiver operating characteristic (ROC) analysis assessed DSS cutoff for ORR prediction. Due to limited tissue and clinical timelines, DST was performed as n=1 biological replicate per patient with technical replicates and appropriate controls on each plate.

• Feasibility and return of results: 21/24 patients (88%) with tissue provided underwent DST; 20/24 (83%) underwent UCSF500 profiling. Nineteen of 25 enrolled patients (76%) completed both DST and genomics and had results reported to the FPM tumor board, exceeding the 60% feasibility target (exact binomial test P < 0.0001; 95% CI 0.5487–0.9064).
• Actionable recommendations: DST yielded actionable treatment recommendations in 21/25 enrolled patients (84%) and in 21/21 (100%) DST-completed patients; genomics identified actionable variants in 5/20 profiled patients (25%), with only 1/5 receiving a cancer-type–matched therapy recommendation. DST identified significantly more options than genomics alone (P < 0.0001).
• Turnaround time: Median DST reporting times were 9 days for hematologic cancers (range 5–17) and 10 days for solid tumors (range 4–23), significantly faster than genomics (median 26.5 days; range 14–63). Rapid DST enabled timely FPMTB discussions and potential therapy initiation.
• DST performance: Using DSS thresholds (effective >10; moderate 0–10), the median number of effective and moderately effective drugs per patient were 21 (range 3–36) and 12 (range 0–32), respectively; all patients had at least three effective agents identified. Overall QC pass rate was 93% (hematologic: 46/48 plates, 96%; solid: 105/115 plates, 91%). Z-prime scores were significantly above the 0.5 cutoff (hematologic P = 0.0045; solid P < 0.0001). Median cell viability at testing was 94% (range 76–98%). Repeated DSTs showed high correlation for DSS and IC50 (P < 0.0001 and P = 0.0005, respectively).
• Genomic landscape: Recurrent alterations included TP53 mutations (30%), CDKN2A/B loss (25%), CBL variants (15%), MYC/MYCN amplifications (5%), PAX3-FOXO1 fusions in alveolar RMS (2/2, 100%), and EWSR1-FLI1 fusions in Ewing sarcoma (2/4, 50%). One AML case had FLT3-ITD (1/2 sequenced AMLs, 50%). Other actionable variants (e.g., SMARCB1 loss; 9p24.1 amplification including PD-L1/PD-L2/JAK2; NRAS p.Q61K) did not match the patients’ cancer types for approved therapies.
• Clinical outcomes: Of 14 patients who received therapeutic interventions post-FPMTB review, 6 received FPM-guided therapy and 8 received TPC. ORR (CR/PR) was 83% (5/6) in the FPM-guided cohort vs 13% (1/8) in TPC (P = 0.0104, Barnard’s test). All FPM-guided patients achieved SD or better; 75% of TPC patients had PD. PFS was significantly longer in FPM-guided patients compared to TPC (P = 0.0037, logrank) and compared to their own previous regimens (P = 0.0001, Cox proportional hazards). PFS ratio ≥1.3 was achieved more often in FPM-guided patients (median 8.5×; range 1.05–48) than in TPC (median 1×; range 0.14–28) (P = 0.0104). Within-patient PFS improved with FPM-guided therapy (P = 0.0313, paired Wilcoxon), but not with TPC (P = 0.9999).
• Predictive validity of DST: Treatment DSS positively correlated with PFS duration (Spearman p = 0.8732, P = 0.0003). Responders (CR/PR) had significantly higher DSS than non-responders (SD/PD) (P = 0.0012). ROC analysis identified DSS > 25 as an optimal cutoff for predicting ORR with AUC = 1.000 and perfect accuracy/precision/recall in this dataset.
• Case example: An AML patient (EV013-AML) with FLT3-ITD had DST-guided selection among FLT3 inhibitors (midostaurin DSS 5.97 vs sorafenib 1.81 vs ponatinib 0), supporting regimen optimization.
• Safety/access considerations influenced regimen choice; when combinations showed comparable DSS, lower-toxicity options were favored. Some repurposing agents (e.g., statins, montelukast) were selected based on DST efficacy and accessibility.

The study demonstrates that integrating ex vivo DST with genomic profiling (FPM) is feasible within a clinically actionable timeframe for pediatric and adolescent patients with relapsed or refractory cancers. DST provided substantially more actionable options than genomics alone and returned results markedly faster, enabling timely tumor board deliberation. Clinical outcomes improved in patients whose treatments were guided by FPM, with significantly higher ORR and longer PFS relative to both TPC and patients’ prior regimens. The correlation of DSS with PFS and ORR underscores the predictive value of the DST platform. These findings align with and extend prior adult hematologic FPM experiences to a pediatric setting that includes both solid and hematologic malignancies. The study also highlights real-world constraints—off-label access, financial barriers, and physician acceptance—that influence the translation of guided recommendations into care. Earlier implementation of FPM may further enhance utility for rapidly progressing diseases. The integration of functional testing into precision oncology supports prioritizing available agents and repurposed drugs to overcome resistance in heavily pretreated pediatric cancers.

This prospective feasibility study shows that functional precision medicine—combining rapid ex vivo DST with genomic profiling—can be delivered within a clinically actionable timeframe for pediatric patients with relapsed or refractory cancers and can guide treatment selections associated with improved objective responses and progression-free survival. DST identified effective therapies for all tested patients and outperformed genomics alone in generating actionable options. The strong association between DSS and clinical outcomes supports DST as a predictive tool. These results support broader adoption and prospective validation of FPM in larger, multi-center studies, earlier integration in the treatment course, systematic assessment of physician acceptance, improved access to targeted agents, and exploration of concurrent sampling from primary and metastatic lesions. Future work may incorporate explainable AI/ML to optimize combination selection and identify multi-omics biomarkers, facilitating routine clinical integration of FPM for rare and high-risk pediatric cancers.

• Small, heterogeneous cohort without a randomized control arm limits generalizability and tumor-type–specific statistical analyses.
• Treatments were at physician discretion; some patients experienced rapid progression precluding implementation of guided options.
• Tissue limitations required n=1 biological replicate per DST and constrained validation experiments.
• Access to off-label targeted therapies was often limited; many guided regimens prioritized readily accessible agents despite high ex vivo efficacy of some targeted drugs.
• In vitro DST may not fully capture tumor microenvironmental dependencies; although mixed-cell PDCs included immune components, this remains a potential limitation.
• Intrapatient heterogeneity between primary and metastatic lesions was not routinely evaluated concurrently.
• Despite rapid DST turnaround (median 9–10 days), some patients’ disease trajectories necessitate even earlier implementation.